AMD Ryzen (Summit Ridge) Benchmarks Thread (use new thread)

[Sweepr]

2600K score 6.51 in Cinebench 11.5 while the FX8350 score 6.92...

So the fact that 3770K is >20% faster than 2600K here doesn't ring any bells for you?

Spoiler

Stock 2600K scores ~6.83-6.92 in multiple websites.

Fixed... thanks.

Please fix it with the correct results now.

 

[rvborgh]

oh my fixed! (hopefully for the last time lol

Stock 2600K scores ~6.83-6.92 in multiple websites.

Please fix it with the correct results now.

 

[PhonakV30]

New Bench from cpchardware

 

[Glo.]

Normalized results put the CPU in this variant on par with 5960X. But with much lower power consumption.

 

[raghu78]

I am surprised AMD let a publication release benchmarks before actual product launch. That too on a chip which is not a final production chip at actual launch frequencies (atleast thats what the google translate seems to indicate). Anyway it looks like 6900k is clearly faster and average IPC is on par with Haswell or slightly lesser. so AMD has to again compete solely on perf/$. I think AMD will have to price 4C/8T at USD 160 - USD 170 , 6C/12T at USD 260 - USD 270 , 8C/16T at USD 350 and 8C/16T flagship SKU at USD 450 - USD 500.

 

[PhonakV30]

Yes Overall is decent.but it comes to Price.

 

[Head1985]

It runs only at 3.3ghz so IPC must be similar to skylake.
Stock 6700k runs at 4ghz whole time-i know this i have 6700k.Thats 21% higher clock than zen.6700k is also 21% faster average.

SO yeah similar IPC like skylake.If zen can be oc from 3.3Ghz to 4.5ghz....

 

[Sweepr]

New Bench from cpchardware

Interesting.

- Application performance 10% above hexa-core 6800K isn't bad at all if it's priced closely
- Looking at 7700K reviews, 8C/16T Ryzen would be ~10-12.5% ahead in application performance based on this chart
- From a technical point of view Intel's 6900K (same core count - 8C/16T) is ~15% faster
- 6900K is only 7.5% faster than 5960X according to Hardware.fr, therefore 2014 Haswell-E octo-core remains faster than this Ryzen
- 6700K still dominates gaming performance - 22% faster than Ryzen, and 7700K is basically out right now (~10% faster than its predecessor)
- Slightly below Core i5 gaming performance, though many of the games tested are very dependant on ST performance
- Power consumption down significantly from Piledriver, but 31W more than 6700K and basically the same as Broadwell-E - and the latter is technically 140W TDP

 

[.vodka]

I am surprised AMD let a publication release benchmarks before actual product launch. That too on a chip which is not a final production chip at actual launch frequencies (atleast thats what the google translate seems to indicate). Anyway it looks like 6900k is clearly faster and average IPC is on par with Haswell or slightly lesser. so AMD has to again compete solely on perf/$. I think AMD will have to price 4C/8T at USD 160 - USD 170 , 6C/12T at USD 260 - USD 270 , 8C/16T at USD 350 and 8C/16T flagship SKU at USD 450 - USD 500.

Doesn't seem like they would let them to. Probably someone got their hands on a sample, and hadn't signed NDA for it.Clock speeds are a little slow, at new horizon Lisa said that 3.4GHz was base clocks for SR... their sample says 3.15/3.4GHz

Still, at these clock speeds, it's certainly a lot more competitive than the FXes they have right now.

 

[PhonakV30]

Interesting.

- Application performance 10% above hexa-core 6800K isn't bad at all if it's priced closely
- Looking at 7700K reviews, 8C/16T Ryzen would be ~12.5-15% ahead in application performance based on this chart
- From a technical point of view Intel's 6900K (same core count - 8C/16T) is ~15% faster
- 6900K is only 7.5% faster than 5960X according to Hardware.fr, therefore 2014 Haswell-E octo-core remains faster than this Ryzen
- 6700K still dominates gaming performance - 22% faster than Ryzen, and 7700K is basically out right now (close to 10% faster than its predecessor)
- Slightly below Core i5 gaming performance, though many of the games tested are very dependant on ST performance
- Power consumption down significantly from Piledriver, but 31W more than 6700K and basically the same as Broadwell-E - and the latter is technically 140W TDP

Wait , Rival of Ryzen Is 6900K not 6700K or 7700K, Why do you mention it ? 6700K is 4 cores/8 Threads while Ryzen 8 cores/16 threads.
Please Just Focus on 6900K or 5960X.no more 6700K or 7700K.

 

[superstition]

No not really, you are not understanding what has happened. by your logic 99% real world apps aren't good benchmarks ( think about that)

Lazy or incompetent developers who haven't bothered to take advantage of old instructions that have been in the market since 2011 and which drastically boost performance is a rebuttal to my point?

Just using a different compiler or two causes a huge performance increase and we're supposed to believe that it makes sense to not just use the better compiler(s)?

It's also a red herring to mention other software because Blender is what AMD chose to use as its benchmark and not all other software needs to be a benchmark or claims to be one.

Now the build that the stilt has built has advanced Vectorization WRITTEN BY A INTEL CODE NINJA. They have likely found a way to vectorize code that was otherwise being executed as scaler FP operations. Now the question is does that advanced vectorization require a operation only found in AVX/AVX2 or can in work on SSE. But either way by Packing those operations together and executing them as one operation you remove pressure on Several bottlenecks that exist in Bulldozers design.

This is fascinating technical talk that dodges what I'm talking about and uses the old "Bulldozer has a bad design" chestnut to get an emotional response (prompting people to forget what we were actually talking about).

Again Blender isn't the problem!. Go build as 2.75 build with AVX2 and watch it performance very close if not the same as 2.75..........

What's the point here? The Stilt's builds offer much better performance on Intel and Piledriver — with current Blender code. 2.75a is a lot slower.

Blender's official builds aren't a problem on Lynnfield but it's not 2009 anymore.

I suppose The Stilt's builds are faster on Skylake because of the deficiencies of the Piledriver architecture.

 

[inf64]

Looks right on the money .
Base 3.15Ghz, max Turbo 3.4Ghz. Assume it runs somewhere in between for full core Turbo (@~3.25Ghz?).
6900K is 3.2Ghz base, 3.5Ghz all core Turbo and 3.7Ghz ST turbo (although even intel lists this SKU as having maximum ST Turbo at 4Ghz on their website).

Frequency wise, Zen ES tested is 8% slower (using best case ST Turbo of 3.7Ghz for 6900K) in ST tests. Similarly it is around ~7.6% slower in all core Turbo clock. Base clock is rather similar -1.5% lower.

On first batch of tests Ryzen ES performed 14.6% slower. Almost all are well threaded SIMD/FP workloads so both Ryzen and 6900K likely ran at all core turbo clocks. Compensating for ~8% lower clock we have Ryzen performing 193.4/ (168.7*1.08)=1.06 or 6% slower than 6900K. 6% is peanuts difference

Second batch of tests were games. Games are not that well threaded so clock is still the king there, although some titles support the multi threading well. The games they tested are CPU hogs tho. Assuming Ryzen ES ran at full turbo all the time(3.4Ghz), and scored 97.3%, If it had ran at 4Ghz like 6700K does, it would have scored ~114% or just 3.5% lower than 6700K. Again, excellent showing.

Last power consumption. Ryzen ES @ 3.15/3.4Ghz based system at full load was drawing 93Watts, 3Watts less than 6900K.

Conclusion :
If AMD was to provide a SKU at the same TDP bracket of 95W and base clocks over 3.4Ghz(like they promised 1 week ago), then we would have a part that will draw less or same as 6900K, perform roughly the same across the board and rock in games. They just need to hit close to 3.8-4Ghz for ST Turbo and give us 3.5/3.6Ghz base clock which would be ~6% higher base than what they demoed a week ago. They can price it to 600-700 dollars and still sell a ton of these.


Edit:
According to this :http://www.overclock.net/t/1619110/cpc-first-unofficial-ryzen-benchmarks
Zen ES has 3.3Ghz all core Turbo so my estimate was off by 3.3/3.25~=1.5% for all core load and ~3% for ST load.

Thanks to Olivon we have another table :

 

[raghu78]

Doesn't seem like they would let them to. Probably someone got their hands on a sample, and hadn't signed NDA for it.Clock speeds are a little slow, at new horizon Lisa said that 3.4GHz was base clocks for SR... their sample says 3.15/3.4GHz

Still, at these clock speeds, it's certainly a lot more competitive than the FXes they have right now.

oh ok. So this does not look like final production chip. Sample running at 3.15/3.4 Ghz and we don't know what the all core turbo is so it still leaves more questions unanswered than answered. With turbo boost 3 Intel core i7 6900k can boost all cores to 3.5 Ghz - 3.7 Ghz depending on workload (INT or FP). Single core turbo boosts to 4 Ghz. Only AVX workloads run at lower than 3.5 Ghz frequency.

http://www.tomshardware.co.uk/intel-core-i7-broadwell-e-6950x-6900k-6850k-6800k,review-33569-9.html

"Again, let’s take a look at the individual core frequencies for the three configurations we tested. Despite a 3.2GHz base clock rate, Turbo Boost pushes the CPU to 3.7GHz in lightly threaded workloads. Our sample managed to maintain this frequency across all cores during the stress test. One core even reportedly hit 4GHz."

So we cannot come to a conclusion on IPC and perf of Ryzen until we have final production chip with base clocks, all core turbo and max core turbo details.

 

[Arachnotronic]

oh ok. So this does not look like final production chip. Sample running at 3.15/3.4 Ghz and we don't know what the all core turbo is so it still leaves more questions unanswered then answered. With turbo boost 3 Intel core i7 6900k can boost all cores to 3.5 Ghz - 3.7 Ghz depending on workload (INT or FP). Single core turbo boosts to 4 Ghz. Only AVX workloads run at lower than 3.5 Ghz frequency.

6900K all core turbo is only 3.5GHz. 3.7GHz is the max single core turbo. 4GHz is only achievable via Turbo Boost Max, which requires software help and isn't really that useful.

 

[inf64]

It is odd that their sample ran at 3.15/3.4Ghz while the OPN suggests 3.5Ghz ST Turbo. Anyways great showing from Ryzen ES.

 

[raghu78]

6900K all core turbo is only 3.5GHz. 3.7GHz is the max single core turbo. 4GHz is only achievable via Turbo Boost Max, which requires software help and isn't really that useful.

dude i am quoting tomshardware test results which actually confirm 3.7 Ghz on all cores and 4 Ghz on single core. If you have other benchmarks and sites please quote them. We don't know if canardpc ran with 6900k with TBM 3.0

 

[KTE]

I am surprised AMD let a publication release benchmarks before actual product launch. That too on a chip which is not a final production chip at actual launch frequencies (atleast thats what the google translate seems to indicate). Anyway it looks like 6900k is clearly faster and average IPC is on par with Haswell or slightly lesser. so AMD has to again compete solely on perf/$. I think AMD will have to price 4C/8T at USD 160 - USD 170 , 6C/12T at USD 260 - USD 270 , 8C/16T at USD 350 and 8C/16T flagship SKU at USD 450 - USD 500.

Pre-release happens all the time, and mostly spot on.

If they've done it now it means launch must be close.

CanardPC is oldskool.

Sent from HTC 10
(Opinions are own)

 

[blublub]

If that "review" is halfway true Ryzen is going to be a decent chip with 3.4ghz base.
If AMD manages 3.6 or even 3.8-4.0ghz base on same Black edition it might even be looking good against Skylake-X

 

[jpiniero]

It's a lot better than expected but not game changing either. You can see why AMD sent it in for another spin... the clocks would need to be higher for it to make sense vs something like the 7700K.

dude i am quoting tomshardware test results which actually confirm 3.7 Ghz on all cores and 4 Ghz on single core. If you have other benchmarks and sites please quote them.

The 3.7 all cores is due to a mobo overclocking setting.

 

[inf64]

If that "review" is halfway true Ryzen is going to be a decent chip with 3.4ghz base.
If AMD manages 3.6 or even 3.8-4.0ghz base on same Black edition it might even be looking good against Skylake-X

Well they promised 3.4+Ghz. So it should be 3.5 or 3.6Ghz base clock which is around 6% higher than what was demoed a ~week ago. I would like to see 3.8Ghz ST Turbo which would put them at 6900K level in games(and elsewhere) according to the canard's results above.

Oh and OP needs to be edited and all the bogus results/images from the top of the opening post replaced with this .

 

[Glo.]

If 8C/16T 3.4/3.7 GHz base chip is 350$, AMD has a killer platform.

 

[Arachnotronic]

dude i am quoting tomshardware test results which actually confirm 3.7 Ghz on all cores and 4 Ghz on single core. If you have other benchmarks and sites please quote them. We don't know if canardpc ran with 6900k with TBM 3.0

On a 6900K? Intel ARK says max single core turbo is 3.7GHz, base clock is 3.2GHz. Puget Systems says all-core turbo is 3.5GHZ.

 

[raghu78]

The 3.7 all cores is due to a mobo overclocking setting.

TBM 3.0 is a feature and with websites testing with that feature turned on we have to wait and see what the actual settings are for 6900k with each site. Similarly for Ryzen final clocks are going to be the real deal. We already know Lisa Su has confirmed base clocks of 3.4 Ghz or higher. The all core turbo and max turbo are the key for Ryzen to make a good impression. No wonder AMD want to keep those details closer to launch.

 

[jhu]

The progression does not suggest that the official Blender build is optimized enough to be considered a credible benchmark for comparing construction core CPUs with Ryzen.

The Stilt's builds call the official builds into question greatly. Ancient AMD processors should not beat much newer ones nor should Lynnfield beat or tie Sandy. In the case of Lynnfield the program would have to either benefit a whole lot from not having SMT enabled or it would not be using recent instructions.

The official Windows builds are done with MSVC 2013. With the official Blender binary for Linux (using GCC) on my FX8350 system running Debian Stretch, I get 1 minute 10 seconds (1371428 samples/s, 83623 samples/s/module/GHz). This is about on par with fully loaded Sandy Bridge. So it could be the tools they're using.

 

[raghu78]

On a 6900K? Intel ARK says max single core turbo is 3.7GHz, base clock is 3.2GHz. Puget Systems says all-core turbo is 3.5GHZ.

Why are you arguing when I quoted tomshardware testing with TBM 3.0. So it depends on how canardpc tested. With TBM 3.0 enabled or off.